7. Acknowledgments
Writing a book is a lot like (we imagine) flying a spaceship too close
to a black hole. One second you’re thinking “Hey, there’s something
interesting over there” and a picosecond later, everything you know
and love has been sucked inside and crushed.
OK, that’s hyperbole, but the point is that books don’t write themselves.
More to the point, books aren’t even just written by the authors. It takes
the combined efforts of a lot of people to extract information from the
chaos. We’d like to hereby issue the following thanks.
To every single beta purchaser of the book and especially the ones who
sent in all those errata posts. You are a fantastic bunch, and we can’t
thank you enough for your belief in the project and your help in making
it a better book.
To the team at the Pragmatic Programmers (especially you, Dave): you
exhibited endless patience, forbearance, and wisdom during the pro-
cess.
Finally, to the authors of all the wonderful frameworks and tools we
highlight in this book: your work is inspiring and we hope that this
book helps shed just a little more light on the work you’ve done.
From Justin Gehtland
To my coauthors: thanks for thinking of me.
My colleagues are an endless font of inspiration and vexation, both
of which help with the creative process. So, thanks to Stu Hal-
loway, Glenn Vanderburg, Neal Ford, and Ted Neward, all of whom
provided various amounts of both.
I keep telling my family that one day I’ll write a book they’d like
to read. At least this one has an interesting cover. Lisa, Zoe, and
Gabe: thanks for putting up with my office hours.
8. A CKNOWLEDGMENTS viii
From Ben Galbraith
Thank you to my family, for all your patience while I spent late
nights and early mornings working on this project. I love you.
My sincere gratitude also goes to my publisher Dave Thomas (who
patiently and gracefully watched this project go from early arrival
to, well, somewhat less than early arrival) and my fellow authors,
Justin Gehtland and Dion Almaer, who made many personal sac-
rifices to get across the finish line.
Finally, I thank all of my peers and colleagues who have taught
me throughout the years. The patience and kindness of nearly
everyone in our industry has always been an inspiration to me.
From Dion Almaer
Ah, acknowledgments. This is the moment where you feel like you
are at the podium and don’t want to forget anyone.
Firstly, I would like to thank my fellow Ajaxians: Ben Galbraith,
Justin Gehtland, Stu Halloway, Rob Sanheim, Michael Mahemoff,
and the entire community that visits and contributes to ajax-
ian.com. This book is really for you, the readers.
Secondly, I would like to thank all of the great technical folk who I
have had the pleasure of working with. This includes buddies from
Adigio, the No Fluff Just Stuff tour, and the general blogosphere.
You know who you are.
Finally, I would like to thank my family, especially my wife, Emily,
who lets me work crazy hours without putting me through guilt
trips. You are my best friend, Em.
9. Chapter 1
Building Rich Internet
Applications with Ajax
This is a book about developing effective web applications. We’re not
going to dance around this issue. Underneath everything else, this
book is about XHTML, JavaScript, CSS, and standards that have been
around for almost a decade now. Not only do we admit this truth, we
embrace it. Just because these standards have been around for a while
doesn’t mean we can’t build something new and exciting out of them.
Technology, like Jello, takes a while to solidify into something tasty and
satisfying.
Ajax (and Web 2.0) represents the maturation of Internet standards into
a viable application development platform. The combination of stable
standards, better understanding, and a unifying vision amount to a
whole that is greater, by far, than the sum of its parts. With Ajax, you’ll
be able to achieve the double Holy Grail: feature-filled user interfaces
and a no-hassle, no-install deployment story.
It wasn’t long ago that Jesse James Garrett coined the term Ajax. When
he first released the term onto the public consciousness, it stood for
Asynchronous JavaScript And XML. It has since, like SOAP before it,
lost its acronym status and is just a word. However, it is an enormously
powerful word. With this single word, Jesse James was able to harness
an industry-wide trend toward richer, install-free web applications and
give it focus.
Naming a thing is powerful. In this case, it’s not powerful enough to
become a movement, though. A spark was still lacking. It was to be
10. A T ALE IN T HREE A CTS 2
provided by an entirely unlikely entity. What follows is the story of
one development team, that spark, and how it changed the way we
approach web software.
1.1 A Tale in Three Acts
Hector is a project manager for a web application development shop.
With a long history of Perl, CGI, ASP, Servlet, and JSP development
under his belt, Hector’s been around the block. For the last year his
team has been building a CRM application for a large Fortune 500 com-
pany with offices all over the world. The application used to be a green-
screen mainframe application; the company wants to take advantage of
the great reach of the Internet to deploy the application to every office.
Hector and his team focus a lot of their energy on the server side of
the application. They have been using one of the modern MVC frame-
works from the Java community to implement the business logic, a
high-performance persistence framework to access the database, and
messaging-based infrastructure to connect to other existing systems.
Yesterday
On the client side, Hector and his team have become masters of CSS.
The look of the pages bends to their will; when the customer wants
rounded corners, they get rounded corners. Rollover colors? That’s
easy. Multiple color schemes? No problem. In fact, Hector and his team
long ago reached a point where they weren’t really worried about the
user interface. See, the Web operates one way: it essentially distributes
static documents. When users want more data, they incur a complete
interface refresh. It isn’t optimal from an efficiency perspective, but it’s
how the Web works, and users have just learned to live with it.
Then, sometime a couple of weeks ago, Hector’s customer came to a
meeting. The customer was usually a polite, accommodating fellow. He
understood the Web, and he understood the restrictions he had to live
with to get the reach of the Internet. In fact, Hector had never seen him
get really angry. Until this meeting.
As soon as he walked in, the team knew something was up. He had his
laptop with him, and he never carried it. As he stormed into the room,
the team glanced around the table: what have we done? The customer
sat down at the table, fired up the laptop, and hammered away at the
keyboard for a minute. While he pounded the keys, he told the team,
11. A T ALE IN T HREE A CTS 3
“Last night, my wife and I were invited to a party at the CEO’s house.”
“Uh oh,” thought the team, “this can’t be good.”
“Well, I certainly jumped at the chance,” he continued. “I’ve never been
before. This project got me on his radar.” (“Double uh-oh,” thought
Hector.) “When I couldn’t figure out how to get there with my city map,
I went to the Internet. I found THIS!” He hissed the last word with
venom and scorn. He flipped the laptop around so the table could see
it. There, quietly couched in his browser window, was Google Maps.
“Why,” he said, through clenched teeth, “can’t I have this?”
Today
Since that meeting, Hector and his team have been rethinking the user
interface. Hector went out to learn how Google could have completely
ignored conventional wisdom and generated such a thing. He came
across an article by Jesse James Garrett describing this thing called
Ajax. He has been digging since then, learning everything he can about
this new way of making Internet applications.
The team has begun reimplementing the UI. They’re using JavaScript
and DHTML techniques to provide a more dynamic experience. Most
of all, they’ve begun taking advantage of a useful object available in
modern browsers called XMLHttpRequest (XHR for short). This handy
little guy lets Hector and his team request and receive fresh data from
the server without reloading everything in the page.
In other words, Hector spearheaded a move from Web 1.0 to Web 2.0.
And his customer is happy again.
Tomorrow
So what comes next for Hector? His team is learning a bunch about
JavaScript, XHTML, and even more about CSS than it ever knew before.
The team is really excited about the results: the user experience is just
like any other application now, except the team doesn’t have to manage
an installer as well as the application itself. But they’ve realized that
there’s a downside to all this.
Now, they are writing a ton of code in JavaScript. It turns out that all
this page manipulation and XHR access requires a lot of real, honest-
to-goodness code. And even though JavaScript looks a lot like Java,
they’ve discovered that it really is a different beast. And now they have
two codebases to manage, test, and maintain.
12. G OOGLE M APS : T HE M ISSING S PARK 4
So Hector is off to find out how to solve these problems. And what
he will see is that most web application development frameworks are
rapidly incorporating Ajax tools into their own suites. Soon, Hector
and his team will be able to leverage Tapestry components, Spring tag
libraries, ASP.NET widgets, Rails helpers, and PHP libraries to take
advantage of Ajax without having to incorporate a second way of work-
ing. The (near) future of Ajax development is total, invisible integration.
And this is exactly what Hector needs.
1.2 Google Maps: The Missing Spark
Google Maps (http://maps.google.com) really ignited the Ajax fire. And
Google was just about the most unlikely candidate to do it. Think
about what made Google an overnight sensation in the first place: bet-
ter search results and the world’s most minimal UI. It was a white page,
with a text box and a button in the middle of it. It doesn’t get any more
minimal than that. If Google had had a soundtrack, it would have been
written by Philip Glass.
When it became obvious that Google was going to enter the online map-
ping space, we all expected something similar: a straightforward, unin-
trusive approach to viewing maps. And this is what we got; just not
the way we expected. Google, through the clever use of XHR callbacks,
provided the first in-page scrollable map. If you wanted to look at the
next grid of map panels, Google went off and retrieved them and just
slid the old ones out of the way. No messy page refresh; no reloading of
a bunch of unchanged text. Particularly, no waiting around for a bunch
of ads to refresh. It was just a map, the way a map ought to work.
Then we clicked on a push pin and got the info bubble. With live text in
it. And a drop shadow. And that was the end of an era. We’ve been told
the same story that you just lived through with Hector again and again.
Somebody’s boss or customer or colleague sees Google Maps and says,
“Why not me?”
As programmers, too, there’s another reaction: “I wish I could work on
that kind of application.” There’s an impression out there that Google
Maps, and applications like it, are rocket science and that it takes a
special kind of team, and a special kind of developer, to make them
happen. This book, if nothing else, will lay to rest that idea. As we’ll
demonstrate in Chapter 2, Creating Google Maps, on page 9, making
web pages sing and dance isn’t all that challenging once you know what
13. W HAT I S A JAX ? 5
tools are available. It becomes even more impressive once you discover
that Google Maps isn’t really proper Ajax; it doesn’t take advantage of
any of the modern asynchronous callback technology and is really just
dynamic HTML trickery.
1.3 What Is Ajax?
Ajax is a hard beast to distill into a one-liner. The reason it is so hard
is because it has two sides to it:
• Ajax can be viewed as a set of technologies.
• Ajax can be viewed as an architecture.
Ajax: Asynchronous JavaScript and XML
The name Ajax came from the bundling of its enabling technologies:
an asynchronous communication channel between the browser and
server, JavaScript, and XML. When it was defined, it was envisioned
as the following:
• Standards-based presentation using XHTML and CSS
• Dynamic display and interaction using the browser’s Document
Object Model (DOM)
• Data interchange and manipulation using XML and XSLT
• Asynchronous data retrieval using XMLHttpRequest or XMLHTTP (from
Microsoft)
• JavaScript binding everything together
Although it is common to develop using these enabling technologies, it
can quickly become more trouble than reward. As we go through the
book, we will show you how you can do the following:
• Incorporate Ajaxian techniques that do not use formal XML for
data transport
• Bypass the DOM APIs themselves for manipulating the in-memory
page model
• Use synchronous calls to the server, which can be powerful but is
also dangerous
• Abstract away the complexity of XMLHttpRequest
It is for these reasons that the more important definition for Ajax is...
14. W HAT I S A JAX ? 6
Ajax: The Architecture
The exciting evolution that is Ajax is in how you architect web applica-
tions. Let’s look first at the conventional web architecture:
1. Define a page for every event in the application: view items, pur-
chase items, check out, and so on.
2. Each event, or action, returns a full page back to the browser.
3. That page is rendered to the user.
This seems natural to us now. It made sense at the beginning of the
Web, as the Web wasn’t really about applications. The Web started
off as more of a document repository; it was a world in which you
could simply link between documents in an ad hoc way. It was about
document and data sharing, not interactivity in any meaningful sense.
Picture a rich desktop application for a moment. Imagine what you
would think if, on every click, all of the components on the application
screen redrew from scratch. Seems a little nuts, doesn’t it? On the
Web, that was the world we inhabited until Ajax came along.
Ajax is a new architecture. The important parts of this architecture are:
• Small server-side events: Now components in a web application
can make small requests back to a server, get some information,
and tweak the page that is viewed by changing the DOM. No full
page refresh.
• Asynchronous: Requests posted back to the server don’t cause the
browser to block. The user can continue to use other parts of the
application, and the UI can be updated to alert the user that a
request is taking place.
• onAnything: We can interact with the server based on almost any-
thing the user does. Modern browsers trap most of the same user
events as the operating system: mouseovers, mouse clicks, key-
presses, etc. Any user event can cause an asynchronous request.
In Figure 1.1, on the next page, we illustrate the new life cycle of an
Ajax page.
1. The user makes an initial request against a given URL.
2. The server returns the original HTML page.
3. The browser renders the page as in-memory DOM tree.
15. W HAT I S A JAX ? 7
R e n d e r e d i n B r o w s e r R u n n i n g o n S e r v e r
H T M L
1 2
3
A D D Y
H E
s c r i p t s /
s e r v l e t s /
4
N K D D
T I T L E L I I V B O I V
p a g e s
6
D S P A N S P A N
I V
5
S P A N S P A N
Figure 1.1: Ajax Page Lifecycle
4. Some user activity causes an asynchronous request to another
URL, leaving the existing DOM tree untouched.
5. The browser returns data to a callback function inside the existing
page.
6. The browser parses the result and updates the in-memory DOM
with the new data. This is reflected on the screen to the user (the
page is redrawn but not “refreshed”).
This all sounds great, doesn’t it? With this change we have to be care-
ful, though. One of the greatest things about the Web is that anybody
can use it. Having simple semantics helps that happen. If we go over-
board, we might begin surprising the users with new UI abstractions.
This is a common complaint with Flash UIs, where users are confronted
with new symbols, metaphors, and required actions to achieve useful
results. Usability is an important topic that we will delve into in Chap-
ter 7, Ajax UI, Part II , on page 122.
Ajax: The Future
Where is Ajax going? What is the future going to hold? This is a vital
question, because Ajax is one of those amorphous terms that seems to
change with the context. Ajax itself is a unifying term for describing
a collection of technologies. We believe that the term itself, as unify-
16. W HITHER N OW ? 8
ing and rallying as it has been, is likely to disappear from the public
consciousness within the next couple of years.
That’s because the technologies you will learn about in this book will
eventually become the substrate of your favorite web application devel-
opment platform. Instead of representing this brave new world of shiny
gadgets and nifty tricks, it will just be how web apps work. Does this
mean that this book is unimportant? Far from it. You need to under-
stand how this works now to get it done, and you’ll need to understand
it in the future to debug your applications. But you probably won’t
think of those apps as Ajax, just as Web apps. And that’s a good thing.
1.4 Whither Now?
The rest of this book will introduce you to the breadth of the Ajax move-
ment. We’ll walk through the conversion of an application to this new
style and provide deep coverage of the enabling technologies behind
Ajax. We’ll introduce you to commonly available toolsets and frame-
works that make seemingly advanced effects as simple as a single line
of code. You’ll get to see what your favorite development platforms are
doing to take advantage of, and integrate with, this new style of devel-
opment.
Most important, we’ll talk a lot about how to use Ajax effectively, prag-
matically, even. That’s because the only thing worse than being left
behind when the train leaves the station is getting on the wrong train.
We intend this book to be a guide through a new and rapidly evolving
landscape. We want to help you find out how, and even if, Ajax can
help your projects. We’re not trying to sell you anything (except this
book). But we believe that Ajax represents a major event, and we want
to be there to help you make the best of it.
But let’s start with the spark that ignited the fire: Google Maps.
17. Chapter 2
Creating Google Maps
For many of us, Google Maps (http://maps.google.com) ignited the Ajax
revolution. While Ajaxian techniques had been creeping into main-
stream websites long before Google Maps, nothing in recent memory
presented commodity browsers with such a visually impressive experi-
ence. Google Maps showed the world that a wide world of potential lay
hidden in the technologies we thought we understood so well.
As we said in Chapter 1, Ajax was initially defined as the intersection
of the XMLHttpRequest object and the usage of XML to update a DOM
tree. However, the current definition of Ajax (and Web 2.0) spans much
more. This chapter demonstrates the underpinnings of Google Maps
and how modern browser-based applications can use nothing but stan-
dard HTML and JavaScript to achieve entirely new kinds of web apps.
The purpose of this chapter is to lay bare the techniques that Google
used to wow us all with Google Maps. What we’ll discover here is fas-
cinating and important; it also might be more than you want to bite off
right now. If so, don’t worry about skipping ahead to the rest of the
book and coming back here later; we won’t mind.
This chapter contains a lot of code. It’s all available online, so you
can download the archives containing all the book’s source.1 Alterna-
tively, if you’re reading the PDF version of this book, just click a link to
get to the file. However, if the file you’re fetching contains HTML, it’ll
probably get rendered by your browser. This is good if you want to see
the running application. If instead you want to see the code, use your
browser’s View Source option.
1 From http://pragmaticprogrammer.com/titles/ajax/code.html
18. R OCKET S CIENTISTS ? 10
2.1 Rocket Scientists?
Shortly after Google Maps launched, entrenched commercial interests
who relied upon the staidness of standard HTML-based web interfaces
to make money were quick to claim that mainstream HTML developers
need not attempt to create web interfaces like Google Maps. The CEO
of Macromedia, maker of the popular Flash browser plug-in, stated
in at least one interview that such non-Flash web interfaces required
the skills of “rocket scientists.” (Ironically, when Macromedia finally
produced a clone of Google Maps in Flash four or five months later, it
failed to function on the two Mac laptops we used to try it out—actually
locking up the browser. Google Maps works just fine on both machines.
We’re actually not anti-Flash; we just found it ironic, that’s all.)
Such statements have added to the general impression many develop-
ers have that creating something like Google Maps is just, well, hard.
In fact, some developers have even felt a little fear and intimidation—
fear that someday soon, they’ll be asked to create something like Google
Maps!
Certainly many of us who have been writing HTML for years might like
to believe that it took a team of rocket scientists to produce a litany of
innovations supporting the technologies behind the Google Maps inter-
face, if nothing else to provide an excuse as to why we haven’t been
writing apps like that all this time. However, we believe all this busi-
ness about rocket science and intimidation is a bit exaggerated.
In fact, after spending ten minutes examining Google Maps a bit deeper,
we realized that, far from being the product of rocket scientists, the
Google Maps interface is actually fairly straightforward to implement.
Perhaps, some might say, easy. Not “same-amount-of-effort-as-a-PHP-
web-form” easy, but we were able to implement something a great deal
like it in about two hours. And this wasn’t just any two hours, mind
you; it was two hours of sitting in a crowded convention center during
a technical conference whilst being interrupted by our friends every few
minutes.
So while there’s no doubt Google has recently hired some of the most
visible computer scientists—perhaps the closest examples of rocket
scientist—like brainpower in our industry, such as Adam Bosworth
(famed Microsoft innovator), Joshua Bloch (famed Java innovator at
Sun Microsystems), and Vint Cerf (famed Internet innovator)—we’re
pretty sure they weren’t involved in the creation of the Google Maps
19. Y OUR O WN G OOGLE M APS 11
The Real Rocket Science
OK, OK we admit—it isn’t easy to create something like Google
Maps. The geocoding features behind the scenes that map
addresses to locations on a map, that normalize a maps fea-
tures against satellite imagery to such an amazing degree that
they can be overlaid on top of each other and look relatively
accurate, and the plotting of routes from Point A to Point B are
all incredibly nontrivial.
However, we maintain that it’s not the geocoding features
of Google Maps that is particularly innovative or impressive.
MapQuest and other software packages have been doing this
kind of work for years. No, what’s impressive about Google
Maps is the web interface on top of the geocoding engine.
And it’s that interface that we find easy, not the geocoding
under the covers.
As our good friend Glenn Vanderburg says, though: “Techni-
cally it’s easy, but the conception of this kind of interface is
the really amazing part, just having the idea and then real-
izing that it could be done. So many things are simple once
you’ve seen that they’re possible.” The take-home lesson is that
Google Maps shows that once you have conceived of your
next great UI idea, you can take comfort in knowing that the
technical solution to implementing it might not be so daunting.
interface. (We should say, though, that we stand in awe of Lars Ras-
mussen and his team for being the brains and fingers behind Google
Maps.) The reality is if we can create an interface like Google Maps in
a couple of hours, imagine what a few capable web developers could do
in a few weeks or a month.
2.2 Your Own Google Maps
In fact, we’ll spare you from putting your imagination to the test. Let
us show you firsthand how you can create your own version of Google
Maps. In the next few pages, we’ll walk you through the creation of
Ajaxian Maps, our own derivative of the big GM. We’ll start out by
explaining how the Google Maps user interface works.
20. Y OUR O WN G OOGLE M APS 12
Figure 2.1: Google Maps
Google Maps Deconstructed
We’re going to break down the elements of Google Maps one by one.
Let’s start out with the most dramatic feature: the big scrolling map,
the heart of the application.
The Map
As you know, the map works by allowing you to interactively move the
map by dragging the map using the mouse. We’ve seen mouse dragging
in browsers for years, but the impressive bit is that the scrolling map
is massive in size, can have the zoom level changed and so forth. How
do they do that?
Of course, the browser could never fit such a large map in memory at
once. For example, a street-level map of the entire world would prob-
21. Y OUR O WN G OOGLE M APS 13
More Than A Million Pixels
We say in “The Map” section that a street-level map of the
world would be about a million square pixels. Actually, that
number’s a wild underestimate. At Google’s highest level of
magnification, a square mile consumes about 7,700,000 pixels.
The Earth is estimated to contain 200,000,000 square miles, but
only 30% of that is land, so let’s reduce the number to 60,000,000
square miles.
Multiplying the number of pixels by the number of square miles
in the Earth produces the mind boggling number of 462 million
million pixels, which at 16.7 million colors (the color depth of
any modern home computer) would consume at least three
times that amount of memory in bytes. Of course, most image
viewing programs have some sort of paged memory subsystem
that views a portion of the image at any one time, but you get
the idea....
ably be about a million pixels square. How much memory would it
take to display that map? For the sake of conversation, let’s assume
that the map is displayed with just 256 colors, meaning each pixel
would consume just 1 byte of memory. Such a map would require
1,000,000,000,000 bytes of memory, or roughly 1 terabyte (1000 giga-
bytes) of RAM. So, simply displaying an <img> element just isn’t going
to work.
What the Googlers do to work around the paltry amount of memory our
desktop PCs have is split up the map into various tiles. These tiles are
laid out contiguously to form one cohesive image. Figure 2.2, on the
next page, shows an example of these tiles. While the size of these tiles
has changed, the current size is 250 pixels square.
The tiles themselves are all laid out within a single HTML div element,
and this div element is contained within another div; we’ll call these
two divs the inner and outer divs, respectively.
We mentioned just a moment ago that the browser couldn’t fit the entire
map image in memory. Of course, dividing a single map into an arbi-
trary number of tiles and then displaying all those tiles at once would
consume an equal amount of memory as the entire image. To compen-
sate for memory limitations, Google Maps virtualizes the grid of tiles
22. Y OUR O WN G OOGLE M APS 14
Figure 2.2: Google Maps Tiles
in memory and displays only the set of tiles that the user can see, in
addition to a few additional tiles outside of the viewing area to keep the
scrolling smooth.
If this whole grid virtualization mishmash sounds a little complex, don’t
worry; it’s fairly straightforward, though it is the most complicated bit
of the UI.
Zoom Level
Another key feature of Google Maps is the ability to zoom in and out,
enlarging or reducing the size of the map, which lets you get a view
of the entire world at one moment and a view of your street the next.
This is actually the simplest of the features to implement. Changing
the zoom level just changes the size of the tile grid in memory as well
as the URLs of the tile images that are requested.
For example, the URL to one of the tiles in Figure 2.2 is as follows:
http://mt.google.com/mt?v=w2.5&n=404&x=4825&y=6150&zoom=3
By changing the value of the zoom parameter to another value, such as
1, you can retrieve a tile at a different zoom level. In practice, it’s not
quite that simple because the grid coordinates change rather a great
deal with each zoom level and they often become invalid.
23. Y OUR O WN G OOGLE M APS 15
Figure 2.3: The Google Maps Push Pin and Dialog
How do they get the zoom level to constantly hover over the map in
a constant position? The zoom level widget is an image embedded in
the outer div, and makes use of transparency to blend in with the map
image.
Push Pins and Dialogs
Other neat-o features are the push pins and dialogs that appear after
a search. Figure 2.3 shows these elements. These are especially cool
because they both include rounded edges and shadows that make them
blend in with the background map in a sophisticated fashion.
We said the zoom level was the easiest feature, and frankly, we were
probably wrong. This is ridiculously easy. The push pins and dialogs
are simply a PNG image. The PNG image format is supported by the
major browsers and supports a nice feature called alpha transparency. alpha transparency
Alpha transparency allows for more than just the simple transparency
that GIF images support; it allows a pixel to be one of 254 different
values between fully transparent and fully opaque, and it’s this gradient
transparency support that allows the push pins and dialog to use a
shadow that blends in with the map.
24. C REATING A JAXIAN M APS 16
Showing these features is simply a matter of positioning images in the
inner div at an absolute position.
Feature Review
There are other features, of course. But we’ll stick to the set of features
we’ve enumerated; we think these represent the vast majority of the
“ooh, ahh” factors. In review, they were as follows:
• The scrolling map: This is implemented as an outer div containing
an inner div. Mouse listeners allow the inner div to be moved
within the confines of the outer div. Tiles are displayed as img
elements inside the inner div, but only those tiles necessary to
display the viewing area and a buffer area around it are present
in the inner div.
• The zoom level: This is an image embedded in the outer div. When
clicked, it changes the size of the grid representing the tiles and
changes the URL used to request the tiles.
• The push pins and dialogs: These are PNG images with alpha
transparency, placed in absolute positions within the inner div.
Now that we’ve deconstructed Google Maps a bit, let’s set about imple-
menting it.
2.3 Creating Ajaxian Maps
Because Ajaxian Maps won’t bother with all of that geocoding mumbo
jumbo, all of our heavy lifting will be in JavaScript. However, we will
use Java to provide some server features and a few image manipulation
tasks.
IE 6, Firefox 1.x, and Safari 2.x Only
We’ve tested this version of Ajaxian Maps in the three major browsers
but haven’t bothered with older versions and more obscure browsers
(sorry, Opera users). It should work on older platforms, but without
testing, we can’t be sure we’ve caught everything.
Step 1: Create a Map
The first step in displaying a map is, err, creating it. While we could
simply steal the wonderful map that Google Maps uses, Google might
25. C REATING A JAXIAN M APS 17
not appreciate that. So, we’ll go ahead and use a map that is explicitly
open source. The Batik project (http://xml.apache.org/batik), an open-
source Java-based SVG renderer, comes with an SVG map of Spain.
We’ll use that.
Because most browsers don’t provide native support for SVG, we’ll need
to convert the map to a bitmap-based format. Fortunately, Batik can
do that for us. One of the nice features of SVG is that it can scale to
arbitrary sizes, so we could conceivably create a huge image for our
map. However, creating truly huge images is a little tricky; because
of memory limitations, we’d have to render portions of the SVG image,
generate our tiles over the portions, and have some sort of scheme for
unifying everything together. To keep this chapter simple, we’ll just
limit our map to 2,000 pixels in width and 1,400 pixels in height. In
order to implement zooming, we’ll also generate a smaller image that
represents a view of the map in a zoomed-out mode.
The following code excerpt shows how to use Batik to convert the map
of Spain into both a 2000x1400 pixel JPG file and a 1500x1050 pixel
JPG file:
File 31 package com.ajaxian.amaps;
import org.apache.batik.apps.rasterizer.DestinationType;
import org.apache.batik.apps.rasterizer.SVGConverter;
import java.io.File;
public class SVGSlicer {
private static final String BASE_DIR = "resources/";
public static void main(String[] args) throws Exception {
SVGConverter converter = new SVGConverter();
// width in pixels; height auto-calculated
converter.setWidth(2000);
converter.setSources(new String[] { BASE_DIR + "svg/mapSpain.svg" });
converter.setDst(new File(BASE_DIR + "tiles/mapSpain.jpg"));
converter.setDestinationType(DestinationType.JPEG);
converter.execute();
converter.setWidth(1500);
converter.setDst(new File(BASE_DIR + "tiles/mapSpain-smaller.jpg"));
converter.execute();
}
}
To compile the code, you’ll need to put the Batik JARs in your classpath
26. C REATING A JAXIAN M APS 18
Figure 2.4: Batik’s SVG Spain Map
(everything in BATIK_HOME and BATIK_HOME/lib) and place the source code
in the following directory hierarchy: com/ajaxian/amaps. Figure 2.4
shows what either map JPG file should look like. You can also replace
the value of the BASE_DIR variable with whatever is most convenient for
you.
Step 2: Create the Tiles
Now that we have a map at two different zoom levels, we need to slice
it up into tiles. This is pretty easy with the nice image manipulation
libraries available in many programming languages. We’ll demonstrate
how to do that with Java here:
File 30 package com.ajaxian.amaps;
import org.apache.batik.apps.rasterizer.DestinationType;
import org.apache.batik.apps.rasterizer.SVGConverter;
import javax.imageio.ImageIO;
import java.io.File;
import java.awt.*;
27. C REATING A JAXIAN M APS 19
import java.awt.image.BufferedImage;
public class ImageTiler {
private static final String BASE_DIR = "resources/";
private static final int TILE_WIDTH = 100;
private static final int TILE_HEIGHT = 100;
public static void main(String[] args) throws Exception {
// create the tiles
String[][] sources = { { "tiles/mapSpain.jpg", "0" },
{"tiles/mapSpain-smaller.jpg", "1"} };
for (int i = 0; i < sources.length; i++) {
String[] source = sources[i];
BufferedImage bi = ImageIO.read(new File(BASE_DIR + source[0]));
int columns = bi.getWidth() / TILE_WIDTH;
int rows = bi.getHeight() / TILE_HEIGHT;
for (int x = 0; x < columns; x++) {
for (int y = 0; y < rows; y++) {
BufferedImage img = new BufferedImage(TILE_WIDTH, TILE_HEIGHT,
bi.getType());
Graphics2D newGraphics = (Graphics2D) img.getGraphics();
newGraphics.drawImage(bi, 0, 0, TILE_WIDTH, TILE_HEIGHT,
TILE_WIDTH * x, TILE_HEIGHT * y,
TILE_WIDTH * x + TILE_WIDTH,
TILE_HEIGHT * y + TILE_HEIGHT,
null);
ImageIO.write(img, "JPG", new File(BASE_DIR + "tiles/" +
"x" + x + "y" + y + "z" + source[1] + ".jpg"));
}
}
}
}
}
Note that to make things interesting, we made our tile size a bit smaller
than Google Maps: 100 pixels square. We chose x0y0z0.jpg as the nam-
ing convention for the tiles, where the zeros are replaced with the x
and y grid coordinates (0-based) and the zoom level (0 or 1; 0 is for the
bigger of the two maps).
Step 3: Creating the Inner and Outer Divs
Now that we have the image tiles, we can start building our map UI.
We’ll start with a simple web page, shown here:
File 32 Line 1 <html>
- <head>
- <title>Ajaxian Maps</title>
- <style type="text/css">
5 h1 {
28. C REATING A JAXIAN M APS 20
Figure 2.5: Humble Beginnings
- font: 20pt sans-serif;
- }
- #outerDiv {
- height: 600px;
10 width: 800px;
- border: 1px solid black;
- position: relative;
- overflow: hidden;
- }
15 </style>
- </head>
- <body>
- <p>
- <h1>Ajaxian Maps</h1>
20 </p>
- <div id="outerDiv">
- </div>
- </body>
- </html>
Figure 2.5 show this page. Pretty simple so far. Let’s get to the good
stuff. The div on line 21 will become what we’ve called the outer div.
The outer div is the visible window into the tiles and will be entirely
contained in the visible space within the browser. The inner div, on
the other hand, will contain all the tiles and be much larger than the
available visible space. Let’s start out by giving it an inner div with
some simple content:
File 33 <html>
<head>
<title>Ajaxian Maps</title>
<style type="text/css">
h1 {
font: 20pt sans-serif;
}
29. C REATING A JAXIAN M APS 21
#outerDiv {
height: 600px;
width: 800px;
border: 1px solid black;
position: relative;
overflow: hidden;
}
#innerDiv {
position: relative;
left: 0px;
top: 0px;
}
</style>
</head>
<body>
<p>
<h1>Ajaxian Maps</h1>
</p>
<div id="outerDiv">
<div id="innerDiv">
The rain in Spain falls mainly in the plains.
</div>
</div>
</body>
</html>
Now we need to make the inner div large enough to contain all of the
image tiles. We could just set a style on the inner div to make it some
arbitrary size, as in <div style="width: 2000px; height: 1400px">, but
we’ll do this via JavaScript. Why? Well, because we’ll implement the
ability to change zoom levels a little later, we know we’ll have to change
the size of the inner div dynamically anyway, so we might as well start
out that way. We’ll use an onload JavaScript handler to initialize the
size of the inner div once we load the page. Check out the code:
File 34 <html>
<head>
<title>Ajaxian Maps</title>
<style type="text/css">
h1 {
font: 20pt sans-serif;
}
#outerDiv {
height: 600px;
width: 800px;
border: 1px solid black;
position: relative;
overflow: hidden;
}
30. C REATING A JAXIAN M APS 22
#innerDiv {
position: relative;
left: 0px;
top: 0px;
}
</style>
<script type="text/javascript">
function init() {
setInnerDivSize( ' 2000px ' , ' 1400px ' )
}
function setInnerDivSize(width, height) {
var innerDiv = document.getElementById("innerDiv")
innerDiv.style.width = width
innerDiv.style.height = height
}
</script>
</head>
<body onload="init()">
<p>
<h1>Ajaxian Maps</h1>
</p>
<div id="outerDiv">
<div id="innerDiv">
The rain in Spain falls mainly in the plains.
</div>
</div>
</body>
</html>
OK, now we’ve got an inner div big enough to display the tiles for the
largest of our two maps. Now we need to add the dragging functionality.
Step 4: Dragging the Map
We’ll implement dragging using three different mouse event listeners.
When the user clicks the mouse in the map area, we’ll use a listener to
indicate that a drag operation has started. Now, if the user moves the
mouse, we’ll use a listener to move the inner div along with the user’s
mouse movements to create the dragging effect. Finally, we’ll use a
listener to turn off the dragging operation when the mouse is released.
The following code demonstrates how we implemented the listeners:
File 35 // used to control moving the map div
var dragging = false;
var top;
var left;
var dragStartTop;
var dragStartLeft;
31. C REATING A JAXIAN M APS 23
function init() {
// make inner div big enough to display the map
setInnerDivSize( ' 2000px ' , ' 1400px ' );
// wire up the mouse listeners to do dragging
var outerDiv = document.getElementById("outerDiv");
outerDiv.onmousedown = startMove;
outerDiv.onmousemove = processMove;
outerDiv.onmouseup = stopMove;
// necessary to enable dragging on IE
outerDiv.ondragstart = function() { return false; }
}
function startMove(event) {
// necessary for IE
if (!event) event = window.event;
dragStartLeft = event.clientX;
dragStartTop = event.clientY;
var innerDiv = document.getElementById("innerDiv");
innerDiv.style.cursor = "-moz-grab";
top = stripPx(innerDiv.style.top);
left = stripPx(innerDiv.style.left);
dragging = true;
return false;
}
function processMove(event) {
if (!event) event = window.event; // for IE
var innerDiv = document.getElementById("innerDiv");
if (dragging) {
innerDiv.style.top = top + (event.clientY - dragStartTop);
innerDiv.style.left = left + (event.clientX - dragStartLeft);
}
}
function stopMove() {
var innerDiv = document.getElementById("innerDiv");
innerDiv.style.cursor = "";
dragging = false;
}
function stripPx(value) {
if (value == "") return 0;
return parseFloat(value.substring(0, value.length - 2));
}
32. C REATING A JAXIAN M APS 24
If you run the code at this point, you’ll now be able to drag that inner
<div> around.
Step 5: Displaying the Map Tiles
The next step requires us to populate our inner div with the map tiles.
Our approach to this will be fairly simple. The scrolling map effect
is achieved by moving an inner div inside of an outer div; therefore,
the tiles we need to display are calculated by determining the current
position of the inner div relative to the outer div and then working out
which tiles are visible in the portion of the inner div that is visible. We’ll
then add those tiles to the inner div.
It turns out implementing this behavior is not terribly difficult. We’ll
create the function checkTiles( ) to do all this and call it from within the
processMove( ) function. processMove( ) is called when the user drags the
map, so by calling it from within, we’ll be able to load our tiles as the
map moves. The following code excerpt shows how we’ve added these
elements to our JavaScript code; for now, checkTiles( ) is just stubbed
out with comments:
File 39 function processMove(event) {
if (!event) event = window.event; // for IE
var innerDiv = document.getElementById("innerDiv");
if (dragging) {
innerDiv.style.top = top + (event.clientY - dragStartTop);
innerDiv.style.left = left + (event.clientX - dragStartLeft);
}
checkTiles();
}
function checkTiles() {
// check which tiles should be visible in the inner div
// add each tile to the inner div, checking first to see
// if it has already been added
}
Now, let’s implement our stubbed-out checkTiles( ) function.
Calculating the Visible Tiles
Calculating the set of tiles that the user can see in the inner <div> is
fairly straightforward. To understand how this works, it will help to
visualize the inner div as a grid where each grid cell is a placeholder of
the tiles that we’ll load. Figure 2.6 illustrates this concept.
33. C REATING A JAXIAN M APS 25
Figure 2.6: The Tile Grid
Because we can’t load all the tiles in the grid up front, we’ll need to
calculate which of these grid cells are visible and load the tiles needed
to fit into these cells. As Figure 2.6 shows, this is accomplished by
calculating which grid cells are visible within the viewport created by
the size of the outer div. In the figure, we see that nine cells are visible
across three rows. Note that those cells that are only partially visible
still count as being visible.
Let’s see how to implement all this behavior we just described. To make
things simple, we’ll encapsulate all of the code to figure out which tiles
are visible in a particular method, which we’ll call getVisibleTiles( ). The
first thing we need to figure out in getVisibleTiles( ) is the position of the
inner div relative to the outer div. This is fairly easy:
34. C REATING A JAXIAN M APS 26
function getVisibleTiles() {
var innerDiv = document.getElementById("innerDiv");
var mapX = stripPx(innerDiv.style.left);
var mapY = stripPx(innerDiv.style.top);
}
The stripPx( ) function, shown earlier, converts the string value returned
by innerDiv.style.left (such as 100px) to a numeric value (say, 100). Now,
we can divide these positions by the size of the tiles to work out the
starting row and column of the tiles. This is just two lines of code:
var startX = Math.abs(Math.floor(mapX / tileSize)) - 1;
var startY = Math.abs(Math.floor(mapY / tileSize)) - 1;
Note that we haven’t yet defined the tileSize variable; we’ll do that glob-
ally (at the top of our JavaScript code), and you’ll see it when we show
the entire page in just a few paragraphs. (Or, you can see it now on
the following page.) The call to Math.floor( ) will round the quotient to an
integer, discarding the remainder (so 1.4 will be rounded down to 1).
This will cause partial tiles to be displayed. Math.abs( ) converts negative
values to a positive number, which in our case is necessary because the
inner div position will nearly always be negative to the outer div, and
because our tile columns/rows are always positive numbers. Finally,
we subtract 1 from the result to make our map load the tiles a touch
early for a smoother effect.
The final bit of calculation is to determine the number of rows and
columns visible in the viewport:
var tilesX = Math.ceil(viewportWidth / tileSize) + 1;
var tilesY = Math.ceil(viewportHeight / tileSize) + 1;
As with tileSize( ), we’ll declare both viewportWidth and viewportHeight as
global variables and show that in just a bit. We use Math.ceil( ), the
opposite of Math.floor( ) (so it rounds the quotient up regardless of the
size of the remainder), to ensure that if any portion of a column or row
is visible, we’ll display it. And, just as we subtracted 1 from the index
of the tiles in the previous lines, we’ll add 1 to the number of columns
and rows to make the scroll effect smooth.
We now have all the data we need to calculate all of the visible tiles
in the viewport plus, as we’ve discussed, a few around the edges that
aren’t immediately visible but will be shortly. Now we’ll build an array
that contains all of the tiles that need to be loaded. To build this array,
we’ll write two for loops, one nested inside the other, that each perform
an iteration for each column and row that is currently visible. Inside
35. C REATING A JAXIAN M APS 27
each loop iteration, we’ll add the column and row number of each tile
to display:
var visibleTileArray = [];
var counter = 0;
for (x = startX; x < (tilesX + startX); x++) {
for (y = startY; y < (tilesY + startY); y++) {
visibleTileArray[counter++] = [x, y];
}
}
return visibleTileArray;
Note that we’re actually creating a two-dimensional array; the value of
each item in our array is another array. We did this because we need
to pass back two values: the column and row index. And now, we’re
done calculating the tiles that are visible in the inner div, and we can
move on and work on the code to actually display them. But first, let’s
review all of the code we’ve written so far:
File 36 function checkTiles() {
// check which tiles should be visible in the inner div
var visibleTiles = getVisibleTiles();
// add each tile to the inner div, checking first to see
// if it has already been added
}
function getVisibleTiles() {
var innerDiv = document.getElementById("innerDiv");
var mapX = stripPx(innerDiv.style.left);
var mapY = stripPx(innerDiv.style.top);
var startX = Math.abs(Math.floor(mapX / tileSize)) - 1;
var startY = Math.abs(Math.floor(mapY / tileSize)) - 1;
var tilesX = Math.ceil(viewportWidth / tileSize) + 1;
var tilesY = Math.ceil(viewportHeight / tileSize) + 1;
var visibleTileArray = [];
var counter = 0;
for (x = startX; x < (tilesX + startX); x++) {
for (y = startY; y < (tilesY + startY); y++) {
visibleTileArray[counter++] = [x, y];
}
}
return visibleTileArray;
}
36. C REATING A JAXIAN M APS 28
Displaying the Visible Tiles
We’ve now coded half of the checkTiles( ) function, which as you may
recall is the function responsible for both calculating the visible tiles
and displaying them. Now, let’s implement the other half of that func-
tion: displaying the tiles.
All we need to do here is iterate through each element of the array of
visible tiles we returned from the getVisibleTiles( ) function and for each
array element add a tile image to the inner div. Here’s the new code for
our checkTiles( ) function:
File 37 Line 1 function checkTiles() {
- // check which tiles should be visible in the inner div
- var visibleTiles = getVisibleTiles();
-
5 // add each tile to the inner div, checking first to see
- // if it has already been added
- var innerDiv = document.getElementById("innerDiv");
- var visibleTilesMap = {};
- for (i = 0; i < visibleTiles.length; i++) {
10 var tileArray = visibleTiles[i];
- var tileName = "x" + tileArray[0] + "y" + tileArray[1] + "z0";
- visibleTilesMap[tileName] = true;
- var img = document.getElementById(tileName);
- if (!img) {
15 img = document.createElement("img");
- img.src = "resources/tiles/" + tileName + ".jpg";
- img.style.position = "absolute";
- img.style.left = (tileArray[0] * tileSize) + "px";
- img.style.top = (tileArray[1] * tileSize) + "px";
20 img.setAttribute("id", tileName);
- innerDiv.appendChild(img);
- }
- }
- }
We start out on line 8 by creating an empty map (map in the JavaScript
sense; a hash that contains key-to-value mappings). We’re going to add
an entry to this map for each visible image; we’ll discuss why we’re
doing this a little later.
On line 9, we start looping through each element in the array we sent
back from getVisibleTiles( ). For each element, we build the name of the
image file that will be loaded in. (If you recall, the file-naming conven-
tion we chose in Step 2 was x0y0z0, where the numbers are replaced
with the index of the tile in the tile grid.) We also use this name as the
key in the visibleTilesMap variable, and on lines 13 and 20 you can see
37. C REATING A JAXIAN M APS 29
that we also use it as the id attribute for each img element that we add
to the inner div. This is so on lines 13 and 14, we can check to see
we’ve already added a given tile to the inner div and, if we have, avoid
adding it again.
Finally, in lines 15 through line 21, we create the <img> element and
add it to the inner div. Note that on line 16 we have to specify the URL
of the image tile. If you have Java installed and executed the code from
Steps 1 and 2 to create your own image tiles, great! Reference them
on line 16, setting the URI to wherever you put them. If not, you can
reference our tiles online.2
You can now enjoy a scrolling map of Spain in your browser! We’ve
placed a copy online at GoogleMaps/step5-3.html. Here’s all the code
we’ve written so far:
File 37 <html>
<head>
<title>Ajaxian Maps</title>
<style type="text/css">
h1 {
font: 20pt sans-serif;
}
#outerDiv {
height: 600px;
width: 800px;
border: 1px solid black;
position: relative;
overflow: hidden;
}
#innerDiv {
position: relative;
left: 0px;
top: 0px;
}
</style>
<script type="text/javascript">
// constants
var viewportWidth = 800;
var viewportHeight = 600;
var tileSize = 100;
// used to control moving the map div
var dragging = false;
2 GoogleMaps/resources/tiles/x0y0z0.jpg, where x0y0z0 should be replaced with the tile you
want to load
38. C REATING A JAXIAN M APS 30
var top;
var left;
var dragStartTop;
var dragStartLeft;
function init() {
// make inner div big enough to display the map
setInnerDivSize( ' 2000px ' , ' 1400px ' );
// wire up the mouse listeners to do dragging
var outerDiv = document.getElementById("outerDiv");
outerDiv.onmousedown = startMove;
outerDiv.onmousemove = processMove;
outerDiv.onmouseup = stopMove;
// necessary to enable dragging on IE
outerDiv.ondragstart = function() { return false; }
checkTiles();
}
function startMove(event) {
// necessary for IE
if (!event) event = window.event;
dragStartLeft = event.clientX;
dragStartTop = event.clientY;
var innerDiv = document.getElementById("innerDiv");
innerDiv.style.cursor = "-moz-grab";
top = stripPx(innerDiv.style.top);
left = stripPx(innerDiv.style.left);
dragging = true;
return false;
}
function processMove(event) {
if (!event) event = window.event; // for IE
var innerDiv = document.getElementById("innerDiv");
if (dragging) {
innerDiv.style.top = top + (event.clientY - dragStartTop);
innerDiv.style.left = left + (event.clientX - dragStartLeft);
}
checkTiles();
}
function checkTiles() {
// check which tiles should be visible in the inner div
var visibleTiles = getVisibleTiles();
39. C REATING A JAXIAN M APS 31
// add each tile to the inner div, checking first to see
// if it has already been added
var innerDiv = document.getElementById("innerDiv");
var visibleTilesMap = {};
for (i = 0; i < visibleTiles.length; i++) {
var tileArray = visibleTiles[i];
var tileName = "x" + tileArray[0] + "y" + tileArray[1] + "z0";
visibleTilesMap[tileName] = true;
var img = document.getElementById(tileName);
if (!img) {
img = document.createElement("img");
img.src = "resources/tiles/" + tileName + ".jpg";
img.style.position = "absolute";
img.style.left = (tileArray[0] * tileSize) + "px";
img.style.top = (tileArray[1] * tileSize) + "px";
img.setAttribute("id", tileName);
innerDiv.appendChild(img);
}
}
}
function getVisibleTiles() {
var innerDiv = document.getElementById("innerDiv");
var mapX = stripPx(innerDiv.style.left);
var mapY = stripPx(innerDiv.style.top);
var startX = Math.abs(Math.floor(mapX / tileSize)) - 1;
var startY = Math.abs(Math.floor(mapY / tileSize)) - 1;
var tilesX = Math.ceil(viewportWidth / tileSize) + 1;
var tilesY = Math.ceil(viewportHeight / tileSize) + 1;
var visibleTileArray = [];
var counter = 0;
for (x = startX; x < (tilesX + startX); x++) {
for (y = startY; y < (tilesY + startY); y++) {
visibleTileArray[counter++] = [x, y];
}
}
return visibleTileArray;
}
function stopMove() {
var innerDiv = document.getElementById("innerDiv");
innerDiv.style.cursor = "";
dragging = false;
}
40. C REATING A JAXIAN M APS 32
function stripPx(value) {
if (value == "") return 0;
return parseFloat(value.substring(0, value.length - 2));
}
function setInnerDivSize(width, height) {
var innerDiv = document.getElementById("innerDiv");
innerDiv.style.width = width;
innerDiv.style.height = height;
}
</script>
</head>
<body onload="init()">
<p>
<h1>Ajaxian Maps</h1>
</p>
<div id="outerDiv">
<div id="innerDiv">
The rain in Spain falls mainly in the plains.
</div>
</div>
</body>
</html>
Cleaning Up Unused Tiles
We’ve got some neat scrolling, but this has one glaring inefficiency. We
add tiles to the inner div on demand, but we never remove the tiles that
are no longer visible. Fortunately, we’ve already done some of the work
to accommodate this feature. If you recall, we created a JavaScript map
named visibleTilesMap in the checkTiles( ) function but never did anything
with it. Now, we’re going to do something.
After we add the image tiles to the inner div, we’ll select all of the img
elements that are present in the inner div, and for each img element,
we’ll check to see whether its id attribute is present in the visibleTilesMap
variable. If so, we know that it’s a currently visible tile and should be
left in the inner div. If not, the <img> is no longer visible and can be
removed. Here’s the additional code in checkTiles( ) to implement this
functionality:
File 38 function checkTiles() {
// check which tiles should be visible in the inner div
var visibleTiles = getVisibleTiles();
// add each tile to the inner div, checking first to see
// if it has already been added
var innerDiv = document.getElementById("innerDiv");
41. C REATING A JAXIAN M APS 33
var visibleTilesMap = {};
for (i = 0; i < visibleTiles.length; i++) {
var tileArray = visibleTiles[i];
var tileName = "x" + tileArray[0] + "y" + tileArray[1] + "z0";
visibleTilesMap[tileName] = true;
var img = document.getElementById(tileName);
if (!img) {
img = document.createElement("img");
img.src = "resources/tiles/" + tileName + ".jpg";
img.style.position = "absolute";
img.style.left = (tileArray[0] * tileSize) + "px";
img.style.top = (tileArray[1] * tileSize) + "px";
img.setAttribute("id", tileName);
innerDiv.appendChild(img);
}
}
var imgs = innerDiv.getElementsByTagName("img");
for (i = 0; i < imgs.length; i++) {
var id = imgs[i].getAttribute("id");
if (!visibleTilesMap[id]) {
innerDiv.removeChild(imgs[i]);
i--; // compensate for live nodelist
}
}
}
Figure 2.7, on the next page, shows what this looks like.
Step 6: Zooming
Zooming is wicked easy; in fact, the hardest bit is just getting a zoom
widget to appear floating above the map. First, we need to create some
kind of image that the user can click on to enable zooming. In Google
Maps, it’s a slider (shown in the margin here); for us, we’ll just create
a simple image that toggles between our two zoom levels. You can use
any image you like; ours is at GoogleMaps/resources/images/zoom.png.
Second, to float the image above the map, we have to properly set the
z-index of our inner div. Browsers support layering elements on top
of each other; the z-index CSS property is used to determine how the
layering occurs. The lower the value, the lower in the layer the element
will appear. Because we want to put our zoom widget above the tile
images, we’ll need to set the z-index of the inner div to 0. The z-index of
the zoom widget then needs to be any value greater than 0 (we use 1).
Now, let’s add the zoom widget. We’ll enclose it in a div, place it inside
the outer div as a peer of the inner div, and we’ll set the z-index proper-
ties appropriately:
42. C REATING A JAXIAN M APS 34
Figure 2.7: Ajaxian Maps!
File 40 Line 1 <body onload="init()">
- <p>
- <h1>Ajaxian Maps</h1>
- </p>
5 <div id="outerDiv">
- <div style="position: absolute; top: 10px; left: 10px; z-index: 1">
- <img src="resources/images/zoom.png"
- onclick="toggleZoom()"/>
- </div>
10 <div id="innerDiv" style="z-index: 0">
- The rain in Spain falls mainly in the plains.
- </div>
- </div>
- </body>
That will give us our floating zoom widget; now we need to create the
toggleZoom( ) function that we referenced on line 8. This will require a
43. C REATING A JAXIAN M APS 35
few minor changes to our code. First, we need to create some sort of
global state that tracks the current zoom level of our map. Second, we
need to reference this state in the various relevant places in our code
(just one, actually).
Let’s start with the global state. We’ll create a variable zoom to track
the current zoom level and while we’re at it add a constant (in the form
of a two-dimensional array) for declaring the two different sizes of the
inner div:
File 40 var zoom = 0;
var zoomSizes = [ [ "2000px", "1400px" ], [ "1500px", "1050px" ] ];
Now, in the name of cleanliness, we’ll change the first line of our init
method from this:
File 38 setInnerDivSize( ' 2000px ' , ' 1400px ' );
to this:
File 40 setInnerDivSize(zoomSizes[zoom][0], zoomSizes[zoom][1]);
There’s just one other place we need to wire in the zoom support: our
checkTiles( ) function, which creates the img elements for the tiles and
gives them their URL. We need to change this hard-coded zoom-level
code:
File 38 var tileName = "x" + tileArray[0] + "y" + tileArray[1] + "z0";
to this:
File 40 var tileName = "x" + tileArray[0] + "y" + tileArray[1] + "z" + zoom;
All that remains is implementing the toggleZoom( ) function, which we’ve
done here:
File 40 function toggleZoom() {
zoom = (zoom == 0) ? 1 : 0;
var innerDiv = document.getElementById("innerDiv");
var imgs = innerDiv.getElementsByTagName("img");
while (imgs.length > 0) innerDiv.removeChild(imgs[0]);
setInnerDivSize(zoomSizes[zoom][0], zoomSizes[zoom][1]);
checkTiles();
}
Nothing too tricky; we swap the value of the zoom variable from 0 to 1,
delete all the <img> elements from the inner div, change the size of the
44. C REATING A JAXIAN M APS 36
Figure 2.8: Ajaxian Maps Zoomed Out
inner div based on the zoom level, and invoke checkTiles( ) to rebuild the
map with the new zoom level’s tiles.
And now, we have zooming in our map application! Cool. The code
for this version is on-line if you need it.3 Figure 2.8 shows the zoom
feature in action, with our map zoomed to the smaller size.
Step 7: Push Pins and Dialogs
The final feature is adding push pins with alpha transparency. When
clicked, these show a dialog that also has alpha transparency. The
3 http://media.pragprog.com/titles/ajax/code/GoogleMaps/step6.html
45. C REATING A JAXIAN M APS 37
hardest part is creating the images.4 These images will not render prop-
erly in IE 6, but see the end of this section for a workaround.
We’re not going to implement a server back end that does searching,
and so on, so just as with zooming we implemented a toggle, we’ll imple-
ment a toggle for our push pin. The graphic for the toggle is available
at GoogleMaps/resources/images/pushpin.png.
We’ll place the push pin toggle right next to the zoom toggle by adding
a new div for it:
File 42 <body onload="init()">
<p>
<h1>Ajaxian Maps</h1>
</p>
<div id="outerDiv">
<div style="position: absolute; top: 10px; left: 10px; z-index: 1">
<img src="resources/images/zoom.png" onclick="toggleZoom()"/>
</div>
<div style="position: absolute; top: 10px; left: 87px; z-index: 1">
<img src="resources/images/pushpin.png" onclick="togglePushPin()"/>
</div>
<div id="innerDiv" style="z-index: 0">
The rain in Spain falls mainly in the plains.
</div>
</div>
</body>
Now we need to implement togglePushPin( ), which, frankly, is a piece of
cake. We’ll just add an absolutely positioned image with a z-index of 1
to the inner div, add an onclick handler to it, and wire that handler to
display the dialog at an absolute position just above the push pin:
File 42 function togglePushPin() {
var pinImage = document.getElementById("pushPin");
if (pinImage) {
pinImage.parentNode.removeChild(pinImage);
var dialog = document.getElementById("pinDialog");
dialog.parentNode.removeChild(dialog);
return;
}
var innerDiv = document.getElementById("innerDiv");
pinImage = document.createElement("img");
pinImage.src = "resources/images/pin.png";
pinImage.style.position = "absolute";
pinImage.style.left = (zoom == 0) ? "850px" : "630px";
pinImage.style.top = (zoom == 0) ? "570px" : "420px";
4 GoogleMaps/resources/images/pin.png and GoogleMaps/resources/images/dialog.png.
46. C REATING A JAXIAN M APS 38
pinImage.style.zIndex = 1;
pinImage.setAttribute("id", "pushPin");
innerDiv.appendChild(pinImage);
var dialog = document.createElement("div");
dialog.style.position = "absolute";
dialog.style.left = (stripPx(pinImage.style.left) - 90) + "px";
dialog.style.top = (stripPx(pinImage.style.top) - 210) + "px";
dialog.style.width = "309px";
dialog.style.height = "229px";
dialog.style.backgroundImage = "url(resources/images/dialog.png)";
dialog.style.zIndex = 2;
dialog.setAttribute("id", "pinDialog");
dialog.innerHTML = "<table height= ' 80% ' width= ' 100% ' >" +
"<tr><td align= ' center ' >The capital of Spain</td></tr></table>";
innerDiv.appendChild(dialog);
}
There’s just one little problem with this new behavior. Do you remem-
ber the image remover code in checkTiles( )? It removes any img element
child of the inner div that has been explicitly added to that function. Of
course, it will clobber our push pin as well, since it is an img child of
the inner div, so we need to modify the function to ignore the push pin:
File 42 var imgs = innerDiv.getElementsByTagName("img");
for (i = 0; i < imgs.length; i++) {
var id = imgs[i].getAttribute("id");
if (!visibleTilesMap[id]) {
if (id != "pushPin") {
innerDiv.removeChild(imgs[i]);
i--; // compensate for live nodelist
}
}
}
We’re done! We’ve implemented all of the features we discussed in the
introduction of this chapter. Let’s wrap up by...err, wait a second.
While Firefox, Safari, and other browsers provide native support for
PNGs with alpha transparency, IE 6 does not. If you’ve been using that
browser to try this sample code, the zoom and push pin buttons as well
as the push pin and dialog itself have looked really awful.
Fortunately, this has an easy (but annoying) fix. Despite not supporting
PNGs out of the box, IE can use some (IE-specific) JavaScript magic
to parse out the alpha channel from a PNG at runtime and display it
correctly. A number of websites document this workaround; in order
to avoid sidetracking our Google Maps story, we’ll just use a JavaScript
47. C REATING A JAXIAN M APS 39
library provided by one of these websites, www.alistapart.com,5 to solve
our problem.
First, we need to include these new JavaScripts in our webpage, which
we’ll do at the top:
File 41 <script language="javascript"
src="resources/js/browserdetect_lite.js"
type="text/javascript">
</script>
<script language="javascript"
src="resources/js/opacity.js"
type="text/javascript">
</script>
Then, because this library requires that the PNGs it fixes be back-
ground images in a div, we need to change our push pin from an img
element to a div, as well as our two toggle buttons, and then finally use
this library to fix all of these divs. We’ll change the toggle button images
to div background images first:
File 41 <body onload="init()">
<p>
<h1>Ajaxian Maps</h1>
</p>
<div id="outerDiv">
<div id="toggleZoomDiv" onclick="toggleZoom()">
</div>
<div id="togglePushPinDiv" onclick="togglePushPin()">
</div>
<div id="innerDiv" style="z-index: 0">
The rain in Spain falls mainly in the plains.
</div>
</div>
</body>
As part of this change, we moved the style attribute settings on the tog-
gle divs into the style sheet we defined at the top of the file (something
we probably should have done anyway):
File 41 #toggleZoomDiv {
position: absolute;
top: 10px;
left: 10px;
z-index: 1;
width: 72px;
height: 30px;
}
5 http://www.alistapart.com/articles/pngopacity
48. C REATING A JAXIAN M APS 40
#togglePushPinDiv {
position: absolute;
top: 10px;
left: 87px;
z-index: 1;
width: 72px;
height: 30px;
}
We now need to add two lines to our init( ) method to use our new IE
transparency library with the toggle divs:
File 41 // fix the toggle divs to be transparent in IE
new OpacityObject( ' toggleZoomDiv ' , ' resources/images/zoom ' )
.setBackground();
new OpacityObject( ' togglePushPinDiv ' , ' resources/images/pushpin ' )
.setBackground();
And finally, we need to reformat the togglePushPin( ) function to use this
new technique:
File 41 function togglePushPin() {
var pinImage = document.getElementById("pushPin");
if (pinImage) {
pinImage.parentNode.removeChild(pinImage);
var dialog = document.getElementById("pinDialog");
dialog.parentNode.removeChild(dialog);
return;
}
var innerDiv = document.getElementById("innerDiv");
pinImage = document.createElement("div");
pinImage.style.position = "absolute";
pinImage.style.left = (zoom == 0) ? "850px" : "630px";
pinImage.style.top = (zoom == 0) ? "570px" : "420px";
pinImage.style.width = "37px";
pinImage.style.height = "34px";
pinImage.style.zIndex = 1;
pinImage.setAttribute("id", "pushPin");
innerDiv.appendChild(pinImage);
new OpacityObject( ' pushPin ' , ' resources/images/pin ' )
.setBackground();
var dialog = document.createElement("div");
dialog.style.position = "absolute";
dialog.style.left = (stripPx(pinImage.style.left) - 90) + "px";
dialog.style.top = (stripPx(pinImage.style.top) - 210) + "px";
dialog.style.width = "309px";
dialog.style.height = "229px";
dialog.style.zIndex = 2;
dialog.setAttribute("id", "pinDialog");
dialog.innerHTML = "<table height= ' 80% ' width= ' 100% ' ><tr>" +
49. C REATING A JAXIAN M APS 41
Figure 2.9: Ajaxian Maps Push Pin and Dialog on IE 6
"<td align= ' center ' >The capital of Spain</td></tr></table>";
innerDiv.appendChild(dialog);
new OpacityObject( ' pinDialog ' , ' resources/images/dialog ' )
.setBackground();
}
And now, finally, we are done. Up until the image transparency bit,
our code was really quite clean and had very little in the way of cross-
browser hacks. Now, unfortunately, it has had to undergo a bit of an IE
makeover, but the consolation prize is that IE 7 natively supports PNG
so all of this may someday be unnecessary.
For review, let’s take a look at our entire page:
